What is the best way to deal with the filter coefficients? - page 3

 
Yuriy Asaulenko:

I take it, judging by the number of coefficients, that you use FIR filters?

I use BIH - there are not many coefficients, but from the normalised form the frequency is more difficult to recalculate. They don't scale that easily.

Yes, they are FIRs. Like them because they have one algorithm for all kinds of filters. BIH is shorter, but now who cares how many hundreds or thousands of multiplications + additions per second the filter does.
 
Alexey Volchanskiy:


It uses a dll from intel and the lapack package.

Is that a bad thing?

In addition to the previous post. Here is a description of the program itself with pictures.


It was based on the idea to make the software better than other programs of similar class (Keny+Goodman). Goodman provided the terms of reference, source codes and explained how to make the software, including the data input interface and calculation algorithms.

 
Alexey Volchanskiy:

I have this question. I am completing an article on filters, the question of coefficients came up . In a nutshell - all parameters and even filter type, low pass filter, high pass filter, bandpass filter, etc. depend on them.


First you need to understand what it's all about.

What do you want to filter, tap water, moonshine, wifi ? ))

 
Sergey Chalyshev:


First you need to understand what you need it for.

What do you want to filter, tap water, moonshine, wifi ? ))

Yes, and another question, how will the coefficients be calculated?
 
Timur Gatin:

Is that a bad thing?


It's a good thing, it would hardly make any sense if it were self-propelled. We need to compare it with Matlab by the length of the coefficient array.
 
Andrey Dik:
Yes, and another question, how will the coefficients be calculated?

I've already written, I'll give you a ready-made set. Perhaps the aforementioned programme will come in handy, I haven't looked at it yet, but I will.
 
Alexey Volchanskiy:

I've already written, I'll give you a ready-made set. Perhaps the aforementioned programme will come in handy, haven't looked at it yet, but I will for sure.

and the ready-made set is derived from what? how is it generated?

will it redraw like hodrick, or only the last bar like mashki?

 
Andrey Dik:

and the finished set is derived from what considerations? how is it generated?

will it redraw like a hodrick, or only the last bar like a mashup?


And I will suggest set parameters later, we can discuss. Conventional FIR filters do not redraw at all if you work on ticks. This last one is re-drawn because it contains dozens or hundreds of ticks even on М1. I will use a tick filter and it is expected to be better than standard moving averages, I will make an analogue of Simple MA that may be used in reality, but not in pure theory.

About the redrawing in the article I'll tell you how it is caused in some cases.

And the sets of coefficients - 15 minutes of computer work. If something does not suit you, we will recalculate it without any problems.

 
Alexey Volchanskiy:

And the sets of coefficients are 15 minutes of computer time. If something is not satisfactory, we will recalculate it no problem.

15 minutes on the computer in Matlab? What's taking so long?
 
Andrey Dik:
15 minutes on a Matlab computer? What's taking so long?

This was from memory, I was counting all possible combinations for myself ) Generally, a filter with a few hundred coefficients is calculated in tens of seconds, for 77 coefficients. - about 2-4 seconds.